---
id: "uci-air-quality-1h"
slug: "uci-air-quality-1h"
name: "UCI Air Quality Chemical Gas Sensor Telemetry"
domain: "Environmental Science & Public Policy"
taskType: "time_series_forecasting"
numRows: 9357
numColumns: 15
targetColumn: "C6H6(GT)"
primaryMetric: "Timestral-123M SQL"
primaryScore: "0.7126"
parquetUrl: "https://huggingface.co/datasets/carlahq/demo-timeseries-benchmark-containers/resolve/main/containers/uci-air-quality-1h/dataset.parquet"
spreadsheetUrl: "https://docs.google.com/spreadsheets/d/1_Yuc_guShuaBemWFyo6OI95s4N7YXASSNsAkrYb3mHo/edit?usp=sharing"
canonicalUrl: "https://carlahq.eu/datasets/uci-air-quality-1h"
---

# Dataset: UCI Air Quality Chemical Gas Sensor Telemetry (uci-air-quality-1h)

> Multivariate environmental time series benchmark capturing 9,357 hourly observations from a chemical multisensor array deployed in an Italian urban road network. The task evaluates multi-target hourly forecasting of critical air pollutants (Benzene, CO, NO2, NOx) across a 168-hour (7-day) operational planning horizon.

## Dataset Overview

| Attribute | Value |
| :--- | :--- |
| **Task Type** | `time_series_forecasting` |
| **Business Domain** | Environmental Science & Public Policy |
| **Target Column** | `C6H6(GT)` |
| **Rows (Instances)** | 9,357 |
| **Features (Columns)** | 15 |
| **Primary Evaluation Metric** | Timestral-123M SQL |
| **Best Benchmark Score** | **0.7126** |
| **Prediction Horizon** | 168 steps |
| **Seasonality** | 24 |
| **Time Frequency** | Hourly |
| **Timestamp Column** | `timestamp` |
| **Entity / ID Column** | `id` |

## Business Decision Context
**Context:** Municipal environmental protection agencies (EPAs) and smart city traffic dispatchers require rolling 7-day forecasts of toxic urban pollutants to schedule industrial traffic restrictions, alert vulnerable populations, and manage urban ventilation corridors.
**Objective:** Predict hourly atmospheric concentrations of ground-truth Benzene (C6H6), Carbon Monoxide (CO), Nitrogen Dioxide (NO2), and Total Nitrogen Oxides (NOx) over a 168-hour horizon H using multisensor array resistance and weather covariates.
**Decision Impact:** Under-forecasting risks severe public health exposure events, WHO particulate exceedances, and delayed municipal smog emergencies; over-forecasting triggers unnecessary low-emission zone traffic shutdowns and costly commercial transport diversions.

## Feature Schema
| Column Name | Data Type | Role | Distinct Values | Missing Values (%) | Description |
| :--- | :--- | :--- | :--- | :--- | :--- |
| `id` | `categorical` | identifier | 1 | 0 | Unique categorical identifier for the urban sensor installation site |
| `timestamp` | `datetime` | timestamp | 9,357 | 0 | ISO 8601 hourly date and time of air sample collection |
| `C6H6(GT)` | `numeric` | target | 407 | 366 | Reference analytical concentration of Benzene measured by standard reference analyzer |
| `CO(GT)` | `numeric` | target | 96 | 1683 | Reference analytical concentration of Carbon Monoxide measured by standard reference analyzer |
| `NO2(GT)` | `numeric` | target | 283 | 1642 | Reference analytical concentration of Nitrogen Dioxide measured by standard reference analyzer |
| `NOx(GT)` | `numeric` | target | 925 | 1639 | Reference analytical concentration of Nitrogen Oxides measured by standard reference analyzer |
| `AH` | `numeric` | known_covariate | 6,683 | 366 | Atmospheric absolute humidity influencing chemical sensor surface conductance |
| `RH` | `numeric` | known_covariate | 753 | 366 | Atmospheric relative humidity percentage |
| `T` | `numeric` | known_covariate | 436 | 366 | Ambient air temperature at measurement site |
| `PT08.S1(CO)` | `numeric` | past_covariate | 1,041 | 366 | Hourly nominal resistance sensor response nominally targeted to Carbon Monoxide |
| `NMHC(GT)` | `numeric` | past_covariate | 429 | 8443 | Reference analytical concentration of Non-Metanic HydroCarbons |
| `PT08.S2(NMHC)` | `numeric` | past_covariate | 1,245 | 366 | Hourly nominal resistance sensor response nominally targeted to Non-Methane Hydrocarbons |
| `PT08.S3(NOx)` | `numeric` | past_covariate | 1,221 | 366 | Hourly nominal resistance sensor response nominally targeted to Nitrogen Oxides |
| `PT08.S4(NO2)` | `numeric` | past_covariate | 1,603 | 366 | Hourly nominal resistance sensor response nominally targeted to Nitrogen Dioxide |
| `PT08.S5(O3)` | `numeric` | past_covariate | 1,743 | 366 | Hourly nominal resistance sensor response nominally targeted to Ozone |

## FEV Benchmark Leaderboard (AutoGluon Time-Series Suite)
Empirical probabilistic evaluation across standard foundation models and baselines:

| Rank | Method / Model | Family | Organization | SQL Win Rate (%) | SQL Skill Score | MASE Win Rate (%) | Median Latency | Execution |
| :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- |
| 1 | **Timestral-209M** | Foundation Model | — | 87.7% | 50.39 | 86.0% | 0.51s | Server (CUDA / PyTorch) |
| 2 | **Timestral-123M (Carla Engine)** | Foundation Model | — | 86.8% | 50.18 | 85.0% | 0.32s | WebGPU (Browser / Sheets) |
| 3 | **TimeRSI** | Closed-Source API | — | 84.8% | 51.24 | 83.8% | 2512s | Server (CUDA / PyTorch) |
| 4 | **t0-gamma** | Foundation Model | — | 84.4% | 49.42 | 81.8% | 4.38s | Server (CUDA / PyTorch) |
| 5 | **Timestral-28M** | Foundation Model | — | 79.2% | 48.42 | 77.0% | 0.11s | Server (CUDA / PyTorch) |
| 6 | **timesfm-3** | Foundation Model | — | 78.5% | 48.99 | 77.0% | 3.41s | Server (CUDA / PyTorch) |
| 7 | **ChakraTS** | Baseline | — | 78.2% | 48.14 | 79.5% | 14.22s | Server (CUDA / PyTorch) |
| 8 | **Lingjiang2.0** | Baseline | — | 73.9% | 47.02 | 71.3% | 8.89s | Server (CUDA / PyTorch) |
| 9 | **Chronos-2** | Foundation Model | — | 70.4% | 47.28 | 68.4% | 0.84s | Server (CUDA / PyTorch) |
| 10 | **timeh2o-v1-tiny** | Foundation Model | — | 69.9% | 45.49 | 67.0% | 0.19s | Server (CUDA / PyTorch) |
| 11 | **Timestral-5M** | Foundation Model | — | 69.0% | 46.19 | 67.7% | 0.04s | Server (CUDA / PyTorch) |
| 12 | **Toto-2.0-2.5B** | Baseline | — | 67.5% | 44.42 | 65.7% | 5.58s | Server (CUDA / PyTorch) |
| 13 | **t0-beta** | Foundation Model | — | 67.4% | 46.65 | 64.5% | 2.29s | Server (CUDA / PyTorch) |
| 14 | **Toto-2.0-1B** | Baseline | — | 66.4% | 44.43 | 65.5% | 2.48s | Server (CUDA / PyTorch) |
| 15 | **TiRex-2** | Foundation Model | — | 65.8% | 45.50 | 63.4% | 0.27s | Server (CUDA / PyTorch) |

## Provenance & Downloadable Artifacts
- **Dataset Parquet File:** [Direct Parquet Download](https://huggingface.co/datasets/carlahq/demo-timeseries-benchmark-containers/resolve/main/containers/uci-air-quality-1h/dataset.parquet)
- **Google Sheets Interactive Demo:** [Open in Google Sheets](https://docs.google.com/spreadsheets/d/1_Yuc_guShuaBemWFyo6OI95s4N7YXASSNsAkrYb3mHo/edit?usp=sharing)
- **Academic Paper / Citation:** [Read Research Paper](https://archive.ics.uci.edu/dataset/360/air+quality)
- **Canonical HTML Page:** [https://carlahq.eu/datasets/uci-air-quality-1h](https://carlahq.eu/datasets/uci-air-quality-1h)